A handwritten Chinese character recognition system based on neural-fuzzy theory
Hsin-Tai Yang, Jue-Wen Lin, Shie-Jue Lee · 2002
In this paper, we propose a stroke-based handwritten Chinese character recognition system, based on neural networks and fuzzy set theory. Our system consists of three main modules that perform stroke extraction, feature extraction, and recognition, respectively. With the introduction of neural net technique and fuzzy set theory, the capability of tolerating transitional and rotational displacements is obtained. The system has been successfully implemented. Two kinds of experiments on similar Chinese characters have been done. One is based on 15 Chinese characters with the same radical. The average recognition rate in this experiment is 97%. The other is based on 23 similar Chinese characters. With 40 training samples for each character, a 91% recognition rate is achieved.